International Journal of Computational Intelligence and Informatics, Vol. 7: No. 4, March 2018 Performance Analysis of Color Images Using Thresholding Techniques for Image Segmentation in Ziehl- Neelsen Sputum Slide Images D. Nithiyapriya I. Laurence Aroquiaraj Department of Computer Science Department of Computer Science Periyar University Periyar University Salem Salem [email protected] [email protected] Abstract - Image segmentation is significant in investigation of clinical images. In this way, viable and accurate techniques are expected to get right diagnosis of quantitative clinical samples. This paper examines the comparison among color thresholding and extensive thresholding methods for Ziehl-Neelsen TB bacilli slide images in sputum samples. Color Thresholding utilizes RGB as input image while Global Thresholding used YCBr as input image. Also we applied Bayesian segmentation to predict the probability of a pixel depicting a ‘TB entity’ by make use of knowledge of Ziehl-Neelsen stain colours and shape size analysis. This technique is testing on N number of different images. The sensitivity and specificity of all tested classifiers is above 99% for the identification of bacillus objects represented by features. Our results may be used to reduce technician involvement in screening for TB, and would be particularly useful in laboratories in countries with a high burden of TB, where, typically, ZN rather than Auramine staining of sputum smears is the method of choice. The experiment results demonstrated the efficiency of the proposed system. 1. INTRODUCTION Biomedical image processing has encountered dramatic extension, and has been an interdisciplinary research field pulling in ability from connected arithmetic, computer sciences, designing, statistics, material science, life sciences and medicine. Sputum slide samples that are smeared with Ziehl-Neelsen method make the acid-fast TB bacilli appears to be red in color while the other cells and organisms retained in blue color as background (Rafikha Aliana A. Raof, Mohd Yusoff Mashor, R.Badlishah Ahmad,). These are important features that could be used to differentiate between TB bacilli and background on Ziehl Neelsen TB slide images. Many studies have been conducted to detect TB bacilli. Veropoulos et al. (1998) and Forero et al. (2003) proposed pattern recognition technique for TB bacilli identification in images of auramine-stained sputum smears using fluorescence microscopes. Yongping et al. (2010) proposed an autofocus algorithm to automate image capturing for TB identification. Recent work by Osman et al. (2010) used moving k-means clustering method to segment TB bacilli from tissue slide images. Such as feature extraction and classification. The main objective of segmentation stage is to partition an image into regions or objects for further analysis Sezgin et al (2004) categorized thresholding techniques into six which are histogram-based, clustering based, entropy-based, attribute similarity, spatial and locally adaptive (Rafikha Aliana A. Raof, Mohd Yusoff Mashor, R.Badlishah Ahmad,). Rafika Aliana raof et al proposed Image segmentation technique for Ziehl-Neelsen sputum slide image is an important factor in extracting tuberculosis (TB) bacilli from the background and sputum cells. Color threshold using Red, Green, Blue (RGB) information has been used widely in many researches. This paper provides new approach to perform thresholding using Hue, Saturation, and Intensity (HSI) information (Rafikha Aliana A. Raof, Mohd Yusoff Mashor, R.Badlishah Ahmad, mohammad, 2010). 2. METHODOLOGY This session was proposed to compare segmentation between color thresholding technique and global Thresholding technique on Ziehl-Neelsen TB slide images. The image acquisition was done using a light microscope with a 40x digital magnification without oil immersion. Before implementation of main ISSN: 2349-6363 177 International Journal of Computational Intelligence and Informatics, Vol. 7: No. 4, March 2018 segmentation process using global and color thresholding methods, the contrast of the images were enhanced using a linear contrast stretching technique and eventually Bayesian theorem has been applied to diagnose the TB. Thresholding method is used to enabling differentiation between TB bacilli and background so that the input images can be easily processed for the next stage which involve with feature extraction and classification. The overview of the proposed approach is shown in figure 1. 2.1. Binarization and Thresholding A binary image is a digital image that has only two possible values for each pixel. A binary image can be processed well than a gray-scale image. The basic idea in Thresholding is to select a threshold (T) to extract an object or several objects with the same value from background. Color Thresholding method selects the threshold by minimizing the within-class variance of the two groups of pixels separated by the Thresholding operator (Rafikha Aliana A. Raof, Zaleha Salleh, Shahiana Izma Sahidan, H, 2008). Thresholding is used for separation of light and dark regions in a given image Figure 6. The basic global threshold, T is calculated as in figure 2 Figure 1. Segmentation process Step 1. Calculate the average gray level in the image and assign it as T. Step 2. Separate the image with respect to T to obtain two groups of pixels: G1 consisting of pixels with grey levels >T and G2 consisting pixels with grey levels ≤ T. 3. Compute the average grey levels of pixels in G1 to give 1 and G1 to give 2. 4. Calculate a net threshold value as follows, Step 3. Compute the average grey levels of pixels in G1 to give and G2 to give . Step 4. Calculate a net threshold value as follows,k (1) Repeat steps 2-4 until the difference in T in successive iterations is less than a predefined limit. Figure 2. Algorithm for Global threshold T calculated 178 International Journal of Computational Intelligence and Informatics, Vol. 7: No. 4, March 2018 This color thresholding method was performed depend on the color information of the image pixels which consists TB, sputum and other objects. RGB color space was detected to be the most opt color paradigm to be implemented with color thresholding technique compared to the other color spaces such as HSI and HSV. The threshold value that was gained from the range of intensities of RGB component was made use of thresholding purpose. If the pixel value where detected not in the intensity range then it will be excluded from the next step process. In equation 1, common algorithm for gray level thresholding was modified as in equation 2. Therefore it can also be applied for color thresholding which have more than one intensity value as in figure 3. (2) Figure 3. Thresholding mapping of input grey level to output grey level 2.2. Global Thresholding Global Thresholding is another thresholding technique used for segmentation purpose. In this method YCBr color space which consists three parts was used as input. The three parts in YCBR color space are luminance. Y, Ry, and By. These parts can be calculated from RGB as in equation 3. (3) In Equation 3, R, G, and B refer to respective red, green and blue components of RGB color model. As for hue, and saturation S, they can be derived as in equation 4 and 5. , (4) S= (5) (6) From Equation 6 refers to a pixel of an input image. The pixels which located in the range will be retained as original. Color and the other background pixels will be assigned as white. This resultant image then will be input image for the next step of image processing. 3. THE COLOR AND GLOBAL THRESHOLDING WITH BAYESIAN APPROACH 3.1. Select the Threshold A brief study on the color information of the mycobacterium and sputum was carried out in order to get the most suitable values to be the selection range of the threshold. The study was carried out on five different types of 179 International Journal of Computational Intelligence and Informatics, Vol. 7: No. 4, March 2018 images, which comprise of normal images, low quality images with low light (under exposed), images which are too bright (over exposed) and blur images. 3.2. Color Thresholding Approach The color thresholding technique was carried out based on the color information of the bacterium to extract it images from the sputum and other objects. This technique specifies the range of RGB intensities for thresholding. The properties of the RGB pixels are being studied to extract the important features from the image. Based on the color information, the color thresholding algorithm should be able to extract the pixels of mycobacterium and reject pixels of other objects. In order to view the important properties of each segment so that necessary features and accurate value of threshold can be obtained from the result, the information is being gathered in a table. In this table, among the features that are noted are the maximum, minimum and average values for each of the RGB components in mycobacterium and sputum respectively. The maximum, minimum and average values for each of the pixels are also noted to extract important characteristic of the RGB values that may be converted into threshold values. Summary of the findings from the study can be visualized in Table 1. Table 1. RGB information in Ziehl-Neelsen TB Bacilli Slide Min Max Average Blue 29 218 133 Red 32 255 158 Green 6 224 141 3.3. RGB Components using thresholding Considering the minimum and maximum values of RGB components, Equations (7), (8), and (9) have been formulated to get the thresholding algorithm. However, the original values have been modified to cater up to 10% of difference. (7) (8) (9) (10) Note that since Equation (10) is the last rule for the whole algorithm, then the final value is either 0 (black pixel) or 255 (white pixel). Where red (x,y) , green (x,y) and blue (x,y) are the pixel value for each of the red and green components respectively.